Reduced natural foods alter bottom-up pressures on American black bears
A Thesis SUBMITTED TO THE FACULTY OF THE UNIVERSITY OF MINNESOTA BY
Spencer James Rettler
IN PARTIAL FULFULLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE
Dr. James Forester
© 2018 Spencer James Rettler
Abstract
Wildlife management is often based on the manipulation of bottom-up and top- down forces. For bear management in North America, food availability (bottom-up) and hunting pressure (top-down) are the primary factors that limit population growth. Whereas seasonal and year-to-year variation in production of fruits and nuts (i.e., primary bear foods) has been widely recognized, long-term trends in natural food production have rarely been reported. Here we compared the current (2015–2017) availability of 18 fruits and nuts, which represent the main foods of American black bears (Ursus americanus) in north-central Minnesota, to what was available during the 1980s. The study area was the same in both time frames, constituting primarily National Forest, where timber harvesting was routine in the 1980s but much less so in recent decades. We hypothesized that forests matured over the 25-year period and consequently produced less fruit for bears. Within each of 12 forest types, we measured the abundance, productivity and biomass (kg/ha) of each fruit-bearing plant species using the same methodology as used in the 1980s. Bootstrapped independent two-sample t-tests identified species-specific changes in food availability between the two sampling periods; generalized linear mixed-effects models were used to quantify changes for groups of fruit-producing species (i.e., summer/fall foods and short/tall shrubs). For all groups of species, the probability of forest stands producing any fruit at all in the recent time period was nearly half (~40%) what it was in the 1980s (~80%). At the landscape scale, we estimated a ~70% decline in biomass availability due to a reduction in young forest types that produce the most fruit, and also a ubiquitous decline in fruit production across most forest types. Our hypothesis that this decline was due to reduced timber harvesting was only partially correct, as we observed the same decline in fruit production within mature forest of the same type and canopy closure, and also along edges of stands with high light penetration. Earlier springs, with potentially greater vulnerability of flowers to frost, and an ongoing invasion of invasive earthworms, which are known to radically affect the soil, may be additional causes for diminished fruit production. Bears will likely need to alter their foraging habits to compensate for lower natural foods as the landscape continues to change. This study demonstrates the complexity of forest management and its unintended effects on wildlife. i
Table of Contents
Abstract……………………………………………………………………………………i
List of Tables…………………………………………………………………………...... iii
List of Figures…………………………………………..……………….……...... iv
Introduction………………………………………………...... 1
Materials and methods………………………………………………………….………....5
Results…………………………………………………………………………….……...14
Discussion…………………………………………………………………………….….20
Conclusion……………………………………………………………………………….25
Tables…………………………………………………………………………………….27
Figures …………………………………………………………………………………...31
References……………………………………………………………………….……….42
Appendix………………………………………………………………………………....48
ii
List of Tables
Table 1: Species groupings for Generalized Linear Mixed Effects models……………..27
Table 2: Abundance results for species groupings……………………………...…….....28
Table 3: Productivity results for species groupings……………………………………..29
Table 4: Biomass results for species groupings…………………………………………30
iii
List of Figures
Figure 1: Map of study area………………………………………………………….….31
Figure 2: Mean species abundance (%) for each forest stand type for the 1980s and
2010s sampling periods…………………………………………………………………..32
Figure 3: Probability of non-zero productivity by sampling period…………………….33
Figure 4: Probability of non-zero productivity in relation to canopy cover…………….34
Figure 5: Year-to-year fluctuations in summer food productivity………………………35
Figure 6: Year-to-year fluctuations in fall food productivity…………………………...36
Figure 7: Mean species biomass by sampling period…………………………………...37
Figure 8: Non-zero biomass in relation to canopy cover………………………………..38
Figure 9: Canopy cover estimates by sampling period………………………………….39
Figure 10: Landscape biomass estimates by sampling period…………………………..40
Figure 11: Landscape biomass estimates by land ownership…………………………...41
iv
1. Introduction
Understanding and measuring the effects of top-down and bottom-up pressures on fisheries and wildlife populations has been an area of interest, as well as a struggle, in ecology for several decades (Hunter and Price, 1992). The abundance and types of foods
(bottom-up) and predators (top-down) can limit populations, with the relative contribution of these two forces varying by the species of interest and the ecosystem it inhabits (Bowlby and Roff, 1986; Estes, 1995; Frederiksen et al., 2006; Power, 1992).
Teasing apart which is the driving force is not always straight forward, particularly when there is within trophic-level competition (Miller, 2008), varying spatial scales
(Gripenberg and Roslin, 2007), or unique behavioral components (Laundré et al., 2014).
This becomes further complicated when the magnitude of these pressures fluctuates over time (Meserve et al., 2003; Ordiz et al., 2013).
For large mammals, humans can become the dominant top-down pressure through hunting (Milner et al., 2007; Ordiz et al., 2013) and vehicular-related mortality (Collins and Kays, 2011). Humans also significantly affect bottom-up pressures via habitat alterations, which can positively or negatively affect wildlife populations (Andrén, 1994;
Vospernik and Reimoser, 2008). Hunting pressure and vehicle-related mortality can affect reproductive strategies (Gamelon et al., 2011), genetic diversity (Allendorf et al.,
2008), and population size and density (Coffin, 2007). Wildlife management agencies often manipulate both hunting pressure and resource availability to achieve desired population sizes and level of offtake (Ahlering and Faaborg, 2006; Bro et al., 2004).
Indeed, it has become commonplace for wildlife agencies to alter habitat to increase food
1 or cover for targeted game species, with the goal of increasing survival, reproduction, and body size –– a strategy first described by Leopold (1933). Oftentimes, though, habitat management has different effects on different species.
Climatic factors also influence bottom-up pressures. It is well established that climate change is having substantial effects on forest net primary productivity, phenology, biogeography, and vegetation composition through increased frequency of natural disturbances (Dale et al., 2001; Overpeck et al., 1990), variable temperatures
(Allen et al., 2010; Hogg and Bernier, 2005; Pastor and Post, 1988), insect outbreaks
(Kurz et al., 2008; Volney and Fleming, 2000), and the spread of pathogens and invasive species (Ayres and Lombardero, 2000; Dukes et al., 2009). In boreal and temperate forests, temperature dictates the beginning and duration of the growing season (Kramer et al., 2000). With warming climate, earlier growing seasons are shifting the timing of resource availability (Cleland et al., 2007; Keyser et al., 2000; Lechowicz, 1995), causing temporally-dependent symbiotic relationships to become disrupted.
In forested environments, different combinations of overstory and understory species compositions, tree densities, and canopy closures will provide distributions of food and cover that favor different wildlife species. Here we focus on how these forest attributes can affect the production of food for an omnivorous, but largely frugivorous big game species, the American black bear (Ursus americanus; henceforth black bear). This forest-dependent species was historically over exploited by hunters; however, it is now increasing numerically and expanding geographically across most of its range in the
USA, Canada, and northern Mexico (Garshelis et al., 2016). The widespread increase in
2 this species is mainly attributable to purposeful control of top-down forces, specifically reductions in legal and illegal hunting and killing of nuisance animals. Also, in some areas, forest cover has expanded (Hall et al., 2003), generally promoting increased vitality of this species.
Food availability affects bear body mass and condition, which greatly influences reproduction (Noyce and Garshelis 1994, Samson and Huot 1995, Costello et al. 2003); this can significantly affect the resilience of the population to hunting (Kontio et al.,
1998). Food-related effects occur in two ways: (1) occasional pulses of abundant or sparse foods may affect age of first reproduction (Garshelis et al., 1998; Howe et al.,
2012) yielding some particularly strong or weak cohorts (Bridges et al., 2011; Garshelis and Noyce, 2008; McLaughlin et al., 1994); and (2) changes in forest composition may affect overall abundance of foods in the long term, altering reproductive output. A prominent example of the latter occurred during a long-term study of grizzly bears (Ursus arctos) in southern British Columbia, Canada. These bears relied heavily on huckleberries (Vaccinium membranaceum) and the super-abundance of this single key food item boosted grizzly reproduction and density, which for two decades were among the highest in North America (McLellan 2011). However, as trees shaded out the huckleberries, bear reproduction plunged (in just 10 years) to one of the lowest in North
America, and bear density declined by 36% (McLellan 2015).
Black bears in Minnesota rely on over a dozen key fruits, which are produced on bushes or small trees in the forest understory from June through October (Garshelis and
Noyce, 2008). Abundance of fruits varies year to year, affecting body condition and
3 reproduction (Noyce and Garshelis, 1994), as well as migration patterns (Noyce and
Garshelis, 2014, 2011). Bears move among different habitats, following week to week changes in the ripening of different fruits (Davis et al., 2006). Habitat-specific abundance (areal coverage) and productivity (fruit production) of the key bear foods was measured on one study site near the center of the Minnesota bear range from the mid-
1980s through the early 1990s. A notable finding was that fruit biomass for bears was especially high in young, regenerating or planted forests (Noyce and Coy, 1990). Annual timber harvest on this study site declined from roughly 70,000–100,000 thousand board- feet (MBF) during the mid-1980s to mid-1990s to 20,000–60,000 MBF during early
2000s to 2015 (United States Forest Service, 2017). Because of this reduced forest cutting, forest composition changed and food-rich stands of young aspen and young pine plantations, which were common in the 1980s, matured and appeared to shade out some of the fruit-producing understory. Over this time period, the bear population declined drastically (Minnesota Dept. Nat. Resources, unpubl. data), possibly due to the combination of reduced food and increased hunting.
Here, we tested whether the food supply for black bears has changed in central
Minnesota since the 1980s and if so, to what extent. We examined whether land-use- related alterations in the composition, structure, and age of a forest have affected food availability, and posit the effects of these changes on the status of a hunted population of black bears. Specifically, we tested whether changes have occurred in food availability
(i.e., abundance, productivity, and/or biomass of fruit-producing plants) within forest types since the 1980s and then examined whether forest composition has changed across the landscape. Finally, we quantified how stand-level food availability and forest 4 composition combine to affect the broad-scale availability of food for bears in northcentral Minnesota.
2. Materials and methods
2.1. Study area
Located in northcentral Minnesota, the Chippewa National Forest (CNF) falls in the transition zone between the boreal forests to the northeast and the temperate forests in the central part of the state. Our study area covered 62,200 ha, dominated by the eastern extent of the CNF (42% of study area); the remainder included some of the George
Washington State Forest (11%), county land (6%), private land (18%), commercial timber industry (8%) and open water (15%) (Figure 1). Both upland and lowland forests were found in this area. Lowlands were dominated by one or more of these species: speckled alder (Alnus incana), black spruce (Picea marina), tamarack (Larix laricina), black ash (Fraxinus nigra), northern white-cedar (Thuja occidentalis), quaking aspen
(Populus tremuloides) and balsam poplar (Populus balsamifera). Uplands forests were dominated by some of these species: quaking and bigtooth aspen (Populus grandidentata), maple (Acer spp.), red pine (Pinus resinosa), paper birch (Betula papyfiera), and balsam fir (Abies balsamea). From 1990 to 2009, 2% of federal land, 4% of state land and 6% of county land was disturbed (e.g., timber harvest, natural blowdown from storms; Garner et al., 2016). From 2010 to 2015, timber harvest activity slightly increased in the CNF (644 ha; 3% of federal land) and in the state and county forests (897 ha; 9% of state/county land). Many lakes, forest roads, and recreational trails can be
5 found throughout the public land. This area was heavily hunted for bears due to the large extent of easily-accessible public land.
2.2. Fruit sampling
We estimated abundance and production of wild fruit-species during early July to late August, capturing the peak fruiting season. The fruit sampling methodology we used during 2015–2017 (n=126 unique forest stands) was patterned after that used by Noyce and Coy (1990), who sampled the same study area during 1984–1989 (n=142 unique stands). We designated 12 forest stand types, based on the species composition and age category of dominant trees: mature aspen (>30 yr), aspen regeneration (5–15 yr), lowland aspen (mixed with balsam poplar), mature red pine (>35 yr), young pine plantations (8–
20 yr), maple, paper birch, recent timber harvests/upland brush (<5 years), black ash, lowland brush, northern white-cedar, and black spruce/tamarack. Initially, we chose stands to sample based on forest stand inventory data derived from public land spatial databases (Itasca County, 2018; Minnesota Department of Natural Resources, 2018;
United States Forest Service, 2018). Before sampling, however, we confirmed the stand type on the ground. We attempted to sample 10 stands of each forest type whenever possible. Furthermore, we attempted to resample the same stands in different years, but had to find alternate ones if the target stand had been harvested. Out of the 126 unique stands, 75 were resampled.
Within each sampled stand, we used flags to temporarily mark the boundaries of
12 circular plots, each with a radius of 3 m. Plots were spaced 30 m apart in a rectangular 6 grid (3 x 4 plots). For small stands, we used closer spacing. For stands with an obvious edge (e.g., abutting a forest road or trail), we positioned one row of plots within 3 m of the edge in order to discern whether fruiting was enhanced by the increased sunlight.
During the 1980s, only interior plots were sampled with the exception of a few stands during the 1988 field season. When comparing the 1980s data to the 2015-17 data, we only used the interior of a stand unless specified otherwise. We measured canopy cover
(%) within each plot, including all vegetation >2 m tall and above the fruit-bearing species, using spherical convex densiometers (Forestry Suppliers, Inc., Jackson, MS,
USA) or by visual inspection. We calibrated visual inspections among observers and with densiometers.
Within each plot, we visually assessed the abundance and productivity of 22 fruit- bearing species consumed by bears (Table 1). We rated abundance on a 0–4 scale based on areal coverage within the plot: 0 = 0%, Trace = 0–1%, 1 = 1–5%, 2 = 5–33%, 3 = 33–
67%, 4 = 67–100%. We used the mid-point percent value for further treatment of these data (e.g., abundance rating of 2 was converted to 19%). We separately rated fruit production, also on a 0–4 scale: 0 = little to no fruit, 1 = below average production, 2 = average, 3 = above average, and 4 = bumper crop. We considered only the number of fruits produced (per plant or per area), not their size or ripeness. To ensure that our ratings matched those of Noyce and Coy (1990), we developed our subjective scale using their fruit yield data (range of values for fruits/m2 corresponding to each subjective rating). To further improve the standardization amongst observers, we would often visually estimate a fruit production rating for a given species in a plot, and then manually count the number of fruit/m2 to confirm that our subjective ratings matched. We also 7 used Noyce and Coy’s (1980) data on the average mass of each fruit (when fully ripe) to convert our subjective ratings to biomass in grams per m2.
2.3. Data analysis
2.3.1. Changes in food availability by forest type
To explore how food availability has changed within forest stand types between the two sampling periods, we first calculated the mean plant abundance, fruit productivity rating, and biomass of each fruit-bearing species among the 12 plots in each stand, and then used the stand average as the sample unit. When a species was not present in a plot, or occurred in only trace amounts, abundance was recorded as zero and the production rating was scored as “not applicable”. Therefore, average production ratings were conditional on the species being present. When a species was absent, biomass was recorded as zero. We removed four species from all further analyses since they were not sampled during all years of both study periods (American spikenard, red-berried elder, nannyberry and hawthorn). We categorized data into two time periods (1980s vs 2010s) for temporal comparisons of abundance and productivity by species and forest type. We also grouped fruits and nuts into categories with similar traits (e.g., short herbaceous or woody shrubs vs. tall woody shrubs or small trees; ripened in early summer vs. late summer or fall foods; Table 1).
We used bootstrapped, independent two-sample t-tests (n=5,000) with a Behrens-
Fisher approach (Efron and Tibshirani, 1994) in Program R (R Core Team, 2017) to test
8 the hypothesis of whether there was any change in food availability (i.e., species abundance, fruit production, and biomass of each fruit-bearing species) within each forest type, across the two time periods with stand as the sample unit (significance level, α =
0.05). This approach was necessary since we were not able to meet the assumptions of normality and equal variance for a typical independent two-sample t-test. For stands that were resampled in different years, their values could not be considered independent, so we averaged these across years. Species that had less than 5 observations in at least one of the decadal sampling periods were omitted from the analysis.
We used Generalized Linear Mixed-Effects Models (GLMM) and Linear Mixed-
Effects Models (LMM) using the package “lme4” (Bates et al., 2014) in Program R to further compare changes in food availability by species grouping (Table 1) for the two time periods. We used average abundance, production rating, and biomass of each species grouping in each stand as response variables. Our predictor variables were the sampling period (1980s vs. 2010s; categorical), average stand canopy cover (percent cover; continuous), upland or lowland stand types (categorical), and deciduous or coniferous stand types (categorical). We included random intercepts for year and unique stand ID to account of non-independence. Due to the nature of the data, the distributions of the production ratings and biomass estimates were zero-inflated. To account for this, we modelled zeros and non-zeros separately using binomial GLMMs with a logit link
(Hall, 2000); models were fit by maximum likelihood using a Laplace approximation
(Raudenbush et al., 2000). These models highlighted factors affecting the probability of a stand having a productivity rating or biomass estimate greater than zero (henceforth referred to as “non-zero”). We then used LMM fit by REML (Bates et al., 2014) to model 9 non-zero productivity ratings and biomass estimates. We used a logarithmic transformation on non-zero values to satisfy the assumption of normality in LMMs
(Verbeke, 1997). We did not need the two-part model approach for abundance since the data had fewer zeros. However, we performed a logit transformation on abundance values
(proportional) to satisfy the assumption of normality of LMMs and added a small constant (1e-4) to zero values (Warton and Hui, 2011). We created one biologically relevant model for each species grouping that included the sampling period (decade), canopy cover, lowland/upland and coniferous/deciduous variables. We generated bootstrapped 95% prediction intervals to quantify uncertainty.
To explore the effect of the edge versus the interior of forests on species abundance and fruit production, we first compared the edge and interior data for the
2015–2017 sampling period, treating edge as a binomial variable and including the same covariates as above (with the exception of sampling period, which was constant). We also compared the 2015–2017 edge data to the 1980s interior data using a similar approach to determine how our current edge data (presumed to represent maximum fruit production) compared to the interior of a forest in the 1980s. The only 1980s edge data were collected during 1988 and were limited to mature aspen (n=11), aspen regeneration (n=10), lowland aspen (n=7), mature red pine (n=8), young pine plantations (n=7), black ash
(n=3) and paper birch (n=7). We tested for differences between the 1988 and the 2015-17 edge data using a similar mixed-effect modeling approach.
Finally, we investigated whether year-to-year variation in fruit productivity during the 2010s differed from that of the 1980s. We plotted yearly mean productivity ratings of
10 common summer and fall foods in stands where these species were the most common and made visual comparisons between the variation in the 1980s to that of the 2010s.
2.3.2. Changes in forest structure and composition
We explored changes in forest structure at the forest-stand level. We used the same bootstrapped, independent two-sample t-test approach from before to discern whether there were differences in canopy cover of forest types by time period. Because canopy cover was estimated differently in each time period (i.e., ocular in the 1980s, densiometer in the 2010s), we needed to calibrate our estimates to be more comparable to the 1980s estimates. We collected paired densiometer and ocular canopy cover estimates in the open and at the edge and interior of forests using two observers at additional plots
(n=15). We fit a simple linear regression of ocular estimate as a function of densiometer estimate. The calibration of densiometer to ocular canopy cover estimates could be modeled by the regression equation: y = -38.94 + 1.44x, where x is the denisometer
2 estimate (%) (F(1,28) = 362, R = 0.9282, P < 0.001).
We also identified changes in forest composition by estimating the proportion of each forest type on the landscape during each sampling period. However, detailed information on forest type composition in the 1980s was not available and forest type information for the 2015–2017 sampling period was only available on public lands (i.e., federal, state and county). To estimate forest composition in the 1980s, we used visual estimates of forest composition collected during weekly flights to locate radio-collared bears. During 1987–88, forest types were identified visually from the air and ranked by 11 dominance (in terms of relative area) within a 500-m radius buffer around each bear location. We only considered bear locations (n=727) in which 50% or more of the buffer overlapped with public land (i.e., federal, state and county) since all fruit sampling was conducted on public land. At the time of the 1980s study, the purpose was to use these rankings to quantify availability of different forest types, since there was no other information on forest composition. Although these are not random points, they represent availability of forest types in the vicinity of locations used by bears and exclude areas that bears might not perceive as available (or useable). Therefore, these buffers will henceforth be referred to as a “bear landscape”. We selected the two highest-ranked forest types within each bear landscape and considered these as equal point samples
(presuming there may have been errors in visually judging which type had the largest area). For the 2010s, we used a comparable approach, except that instead of visually identifying forest types from an airplane, we overlaid the 727 bear landscapes used in
1987-88 on current stand inventory layers for public land and extracted forest type information (two dominant types) using ArcMap 10.5 (no data were available for private land). There was no acceptable spatial layer for lowland brush; however, this land cover type was unlikely to have changed so we assumed that it comprised the same proportion of the landscape as in the 1980s. Using these samples, we calculated the proportion of each forest type on a bear landscape.
2.3.3. Changes in food availability on a bear landscape
12
To determine whether the total amount of food available on a bear landscape had changed since the 1980s, we needed to extrapolate our mean biomass by forest type across a bear landscape for both time periods. Using the estimated proportion of each forest type on an average bear landscape, we multiplied this by the total area within a
500-m radius buffer (78.5 ha). We then multiplied the estimated area of each forest type by the mean total biomass of fruit produced within each forest type for the corresponding sampling period. Standard errors were calculated based on the variation amongst the total stand biomasses. We also compared the estimated proportions of each habitat type for the
2010s obtained through this sampling technique to the true proportions obtained using
ArcMap to determine how our sampling compared to what was actually on the landscape.
We further explored landscape fruit biomass availability based on the forest composition of different public land ownerships during the 2015-17 sampling period.
Compared to federal lands, state and county land had more timber harvest activity, and consequently more young forests; however, our sample size for state and county forest stands was too small to determine whether average species biomass within a specific forest type significantly varied between different public land ownerships. Instead, we made the simplifying assumption that average species biomass did not vary by ownership for each forest type, and overall availability of bear foods varied only by how much of each forest type was on the landscape. We calculated the proportion of each forest type on federal and state/county lands using the stand inventory maps and then multiplied that proportion by the total area of public land within the study area (36,692 ha). Finally, we multiplied the average fruit biomass within each forest type by the respective area of each forest type on federal and state/county land. 13
2.3.3. Frost vulnerability over time
We explored whether warmer and more variable winters and springs in the 2010s could have affected productivity by causing earlier flowering, which could be more vulnerable to a late frost. Daily minimum and maximum temperatures were obtained from NOAA (NOAA, 2018) for the specific years of fruit sampling. We obtained flowering dates from the USA National Phenology Network (NPN, 2018) and corresponding accumulated growing degree days (AGDD) from 2010-2016 for chokecherry (Prunus virginiana) and red-osier dogwood (Cornus sericea) within the study area. While we were limited to these two species, they were a potential proxy for summer and fall foods, respectively. We used the median AGDD for 2010–2015 to predict dates of flowering. We compared how many times the daily minimum temperature dropped below 0°C (i.e., potential frost events) after these predicted flowering dates during the 1980s versus 2010s.
3. Results
3.1. Food availability within forest types
3.1.1. Plant abundance
Five fruiting species declined in average abundance between the two decadal sampling periods; however, the difference in mean abundance estimates were ≤5% for
14 four of these species. The most notable decline was for sarsaparilla in aspen regeneration
(12.3% decline, P < 0.001) and paper birch (10.6% decline, P <0.001; Figure 2).
Abundance of beaked hazel (in mature aspen, lowland aspen and young pine plantations) and downy arrowwood (in mature red pine) increased between the sampling periods (P <
0.05). The largest increase was for beaked hazel in mature aspen (13.3%, P < 0.05; Figure
2).
Notwithstanding these species-specific changes, the abundance of plant species groups (Table 1) that produced fruit or nuts (during the summer, fall and on short or tall shrubs) did not change between the 1980s and 2010s (Table 2). The abundance of most fruit-producing plants was lower in lowland forests than in upland forests in both sampling periods (P < 0.001, Appendix Figure 1). Only short shrub abundance declined with increasing canopy cover (P < 0.001, Appendix Figure 2). Furthermore, edge plots during 2015–17 had a higher abundance of tall shrubs and species that produce fall food than interior plots (P < 0.05). Short shrub abundance was lower on edges of a stand in the
2010s than in the interior of a stand in the 1980s (P < 0.05); however, the abundance of all species groups did not vary between the 2015–17 edge plots and the 1988 edge plots.
3.1.2. Fruit production
Ten species out of 18 declined in fruit production between the 1980s and 2010s; however, 130 of the 204 unique combinations of species and stand type could not be analyzed for fruit production due to small sample size or the absence of the plant in many stands. Pin cherry in pine plantations had the largest decline in productivity of any 15 species (1 point on 0–4 scale; P < 0.05) while the productivity of sarsaparilla declined in the most forest types (6 of 12, with the largest decline in black ash [-0.40; P < 0.01]).
Beaked hazel productivity declined in five forest types (P < 0.001) while raspberry and juneberry declined in four (P < 0.05). Productivity of beaked hazel and juneberry declined most in aspen regeneration, whereas raspberry fruiting declined most in young pine plantations (P < 0.05). Productivity did not increase for any species from the 1980s to 2010s.
The probability of a stand having some fruit production (i.e., non-zero) declined across all species groups (Table 1) from the 1980s to 2010s (P < 0.001). Short shrubs
(86.6%), tall shrubs (83.7%), summer food species (87.4%), and fall food species
(80.0%) had a higher probability of producing fruit in the 1980s than in the 2010s
(43.6%, 39.5%, 38.8%, 43.4% respectively; P < 0.001, Table 3 and Figure 3). However, when fruits were present, there was no difference in production ratings (Table 3). Upland forests had a higher probability (80.7% and 79.9%) of having tall shrubs and fall food species produce fruits than lowland forests (40.0% and 32.1%; P < 0.001, Table 3 and
Appendix Figure 3). Increased canopy cover reduced the probability of short shrubs and summer food species producing fruit (probability of fruit production declined by 0.20 logit units with each 10% increase in canopy; P <0.001; Figure 4 and Table 3). When fruits were present, their productivity scores declined with increasing canopy (mean productivity for short shrub and summer food species declined 0.10 log units with each
10% increase in canopy; P < 0.001; Appendix Figure 4 and Table 3).
16
Edges and interior sections of a stand had a similar probability of producing some fruit. However, when fruits were present, edges produced more fruit for summer food species, fall food species, short and tall shrubs (P < 0.05). The edges of stands in the
2010s had a lower probability of producing fruit than the interior of stands in the 1980s for all groupings of plants (P < 0.001), but, for stands that produced fruits, there was no difference in fruit production ratings on the edges of stands in the 2010s versus the edges in the 1980s.
Year-to-year variation in productivity was greater for summer food species than fall food species, and appeared to be greater in the 1980s than the 2010s for common species (Figures 5 and 6). However, we could not compare these statistically, and the number of sampled years in the 1980s was twice that of the 2010s. Notably, the lowest summer production year of the 1980s (1989) was roughly equivalent to all the 2010s, and the lowest fall food production year of the 1980s (1985) was similar to the lowest of the three years in the 2010s (2015).
3.1.3. Fruit biomass
Half (9 of 18) of the fruiting species sampled declined in biomass from the 1980s to 2010s (Figure 7). The greatest declines in fruiting biomass were raspberry (-87.8 kg/ha, P < 0.05) and blackberry (-48.7 kg/ha, P < 0.01) in pine plantations, and chokecherry (-36.5 kg/ha, P < 0.05) in lowland aspen. Biomass of beaked hazel and sarsaparilla declined in most forest types (4 of 12). Both of these experienced the greatest
17 decline in aspen regeneration (-21.5 kg/ha for beaked hazel, P < 0.01; -3.5 kg/ha for sarsaparilla, P < 0.001). No species increased in biomass between the two decades.
The overall difference in biomass of fruits between the 1980s and 2010s was due to the lower probability of stands producing any fruit in the 2010s (P < 0.001, Table 4 and Appendix Figure 5), not on the biomass of fruits in stands when present. As canopy cover increased, total fruit biomass declined, and the probability of zero biomass increased for summer food species and short shrubs (P < 0.001 for all comparisons, Table
4, Figure 8 and Appendix Figure 6).
3.2. Forest structure and composition
3.2.1. Canopy structure
Mean canopy cover estimates of the interior of mature stands increased between the sampling periods for mature aspen, black ash, lowland aspen, and mature red pine and declined for upland brush (P < 0.01, Figure 9). The largest increase in canopy cover was for lowland aspen forests.
3.2.2. Forest composition
A major shift occurred in the age of aspen over the three decades (Appendix
Figure 7). The availability of mature aspen in the landscape used by bears on public lands increased from 28% in the 1980s to 49% in the 2010s, whereas young aspen declined from 18% to <1%. Pine plantations showed the same trend: young plantations comprised
18
3.4% of the landscape in the 1980s versus <1% in the 2010s, whereas mature pine increased from 4% to 8%.
3.3. Bear landscape food availability
Our method of estimating the composition of the landscape around bear locations yielded a higher proportion of mature aspen, spruce/tamarack and upland brush than the true proportions of these forest types in the defined study area during the 2010s
2 (X 11=1836.3, P < 0.001, Appendix Figure 8). Within a bear landscape, the reduction of young aspen and young pine on the landscape yielded an estimated loss of >2,000 kg of bear foods, and the reduction in birch caused an additional loss of >700 kg from 1980s to
2010s (Figure 11). All forest types combined, there was an estimated 70% decrease in the total biomass for the portions of the study area used by bears.
We estimated that state and county lands had a higher proportion of regenerating aspen, upland brush, and pine plantations than federal lands (Appendix Figure 9).
Estimated fruit biomass production averaged 17.1 kg/ha on federal lands versus 26.5 kg/ha on state and county lands (Figure 12).
3.5. Flowering dates and frost vulnerability
There was no difference in the estimated flowering dates in the 1980s and 2010s for either species (Appendix Table 1). There was no difference between the two sampling
19 periods in the number of potential frost events for chokecherry. Red-osier dogwood flowers did not experience any potential frost events in either sampling period.
4. Discussion
We observed a large (70%) reduction in natural food availability for bears within forest stands across the landscape from the 1980s to 2010s. The large magnitude of this decline was likely sufficient to affect the biology and behavior of bears. The best fruit production years in the summer and fall of the 2010s were similar to food failures in the
1980s. This decline in food availability over three decades occurred across many forest types and was driven by two factors: (1) changes in composition of the landscape, with a reduction in young, productive forest types, and (2) reduced productivity in stands of the same age and type. For the most part, the decline was not driven by a reduction in abundance of fruit-producing plants. Most fruit-producing plants in northern Minnesota were large, woody shrubs or small trees, which would not change much in abundance without a significant disturbance. Conversely, short herb and shrub species whose vegetative structures die back and regrow each year showed a greater change in abundance. Most notable of these were sarsaparilla and raspberry.
Sarsaparilla is typically the first fruit-bearing species to ripen in the summer and was one of the most common berries found in bear scats during the 1980s (Garshelis and
Noyce, 2008). Since then, its abundance, production and fruit biomass have declined precipitously in many of the forest types in which it was found. Likewise, raspberry was
20 the single largest contributor to overall fruit biomass in younger forests in the 1980s
(Noyce and Coy 1990), but has drastically declined over the past few decades. Thus, bears in this area have lost two early summer staples in their diet. Additionally, several important late summer and fall foods have declined, notably blackberry, chokecherry, and beaked hazel. During the 1980s, blackberry and chokecherry contributed the most late summer fruit biomass in upland and lowland forests, respectively (Noyce and Coy 1990).
Now these species barely contribute to the food availability in aspen regeneration, pine plantations and lowland aspen. The reduction in beaked hazel production is especially striking because this is the most important fall food across much of Minnesota, and so its abundance can strongly affect bears’ vulnerability to hunting (which is done using bait) and their condition before entering dens (Garshelis and Noyce, 2008; Noyce and Coy,
1990).
We found support for our initial hypothesis that a prolonged reduction in timber harvest activity in the Chippewa National Forest led to a change in forest structure (i.e., increased canopy cover), an increase in mature forest types on the landscape, and consequently a reduction in food availability. Most of the mature forest stands were >40 years old; therefore, they were likely in the understory reinitiation phase of stand development post timber harvest (Huang et al., 2013; Oliver, 1980). Competition for sunlight is high during this time and understory tree species begin to fill in new gaps as the dominant tree species naturally thins (Huang et al., 2013). Sugar maple (Acer saccharum) is a common late-successional species in mesic sites in the Upper Great
Lakes region (Heinselman, 1954) and was often present in the understory and mid- canopy of mature aspen and paper birch stands on our study area. We observed similar 21 cases with red maple (Acer rubrum) in the mature red pine mid-canopy, and new cohorts of black ash filling the mid-canopy in lowland aspen and black ash stands. This likely explained the increase in canopy cover of these mature forest types since the 1980s, which lowered the productivity of many of the shade-intolerant fruiting species.
However, this relationship does not explain the large reduction in food availability in young forests, as well as why most of the stands we surveyed in the 2010s did not produce any fruit or nuts. Furthermore, while the edge of forests did have higher productivity than the interior during the 2010s, due to greater light penetration, the probability of any fruit production on the edges of forests during the 2010s was still much lower than the interior of forests during the 1980s.
To investigate why even young forests and forest edges with plentiful light did not produce any food in the 2010s we tested the hypothesis that climate warming shifted the phenology of flowering to earlier in the spring and increased their exposure to damage from a later frost. While we know that killing frosts do occasionally destroy berry crops in Minnesota, we did not find any evidence to indicate that this was more common in the 2010s than the 1980s. However, we had flowering dates for only two species, and we only examined temperature data. Frost events are dependent on the duration of below-freezing temperatures and humidity, particularly at the forest floor
(Inouye, 2001). This is further influenced by the forest structure: forests with a more open canopy are more prone to frost (Ågren, 1988). Spatio-temporal variability of flowering on the landscape compounds the difficulty in assessing the changing effects of frost events to plant reproduction. More careful studies focused specifically on the effects of frost have shown an increase in mortality of wildflowers in recent years due to a climate 22 change-related shift in phenology (Inouye, 2008). Without detailed historical and contemporary information on plant phenology in relation to weather, we cannot conclusively determine whether the flowering dates have changed or whether the frequency of frost events affected fruit production.
Another hypothesis is that invasive European earthworms may be affecting fruit production. There is clear evidence that earthworms have increased in the CNF, and their infestation of the forest was still in process decades ago; because the worms were introduced from fishing bait, their colonization is strongly associated with the many fishing lakes, cabins, and road access in the CNF (Holdsworth et al., 2007a). In fact, the
CNF is the most road-dense national forest in the country, and also has the most lakes
(United States Forest Service, 2018), and our study site has a particularly high density of small-medium-sized lakes accessible for fishing (Figure 1). It is known that earthworms have a major effect on the physical, chemical, microbial, and mycorrhizal composition of the soil, and this significantly reduced abundance and diversity of understory plant species in the CNF and other northern temperate forests (Bohlen et al., 2004; Hale et al.,
2006; Holdsworth et al., 2007b). The large decline in abundance of wild sarsaparilla in the CNF may be explained by the rarity of this species in areas heavily-infested by earthworms (Frelich et al., 2006; Holdsworth et al., 2007b). We have not seen studies relating abundance of invasive earthworms to fruit production, but this seems like a logical extension given the reduction in soil nutrients such as nitrogen and phosphorous, and consequent stress to rooting systems following an earthworm invasion (Dobson et al.,
2017; Frelich et al., 2006). If this hypothesis is correct, it may mean that the decline in bear foods coincides with earthworm infestations, which are much more widespread in 23
Minnesota forests than just the CNF. Also, as noted by Frelich et al. (2006), the relationships between earthworms and forest productivity may be exacerbated by other agents, such as deer browsing and climate change, and climate warming is likely to exacerbate the earthworm invasion in closed canopy forests (Eisenhauer et al., 2014).
This reduction in food resources could have profound impacts on bears’ body condition, reproduction, cub mortality (starvation), adult mortality (hunting vulnerability), and ultimately the carrying capacity of the landscape. In an analogous case on an island in Washington state, heavy logging promoted a rich pulse of fruiting, which supported a vigorous black bear population; when logging ceased, the bear population crashed (Lindzey et al. 1986). Numerous other studies have demonstrated how a paucity in food resources in particular years have negatively impacted body condition and reproductive potential of north-temperate bears (Costello et al., 2003; Elowe and
Dodge, 1989; Hertel et al., 2017; Noyce and Garshelis, 1994; Seger et al., 2013).
When faced with variable natural foods, bears in Minnesota and elsewhere have exhibited adaptability by altering their foraging habits (Clark et al., 1994; Costello and
Sage, 1994; Hashimoto et al., 2003; Hertel et al., 2016). However, this may also prompt bears to rely on risky behaviors to obtain necessary calories. When natural foods are scarce, bears often seek calorically-dense anthropogenic foods such as garbage and crops, bringing them into proximity to humans and increasing their chance of being killed
(Beckmann and Berger, 2003; Ditmer et al., 2016; Kavčič et al., 2015; Lamb et al.,
2017). For bear populations that are harvested, there is strong evidence that years of poor food availability results in more visitations to hunter’s baits and, consequently, increased
24 hunter success (Clark et al., 2005; Noyce and Garshelis, 1997; Obbard et al., 2014). In
Minnesota, hunting success has increased dramatically in recent years, which may in part be due to bears’ increased reliance on hunters’ baits (Garshelis and Tri, 2018). Thus, there is an interaction between top-down and bottom-up pressures, which may amplify the negative impacts on the population.
5. Conclusion
Wildlife agencies tend to manage bear populations under the assumption that top- down forces control population size, possibly taking into account periodic food failures but not long-term trends in food availability. Measuring food availability is tedious and time-intensive, and requires a long-term dataset to discern trends. It is not unusual for agencies managing bears to conduct surveys of a few key mast species (e.g., Clark et al.,
2005), but we are unaware of surveys like ours involving such a large diversity of summer and fall bear foods. Moreover, our goal was not just to look for differences in availability over time, but to more deeply examine where and why these differences occurred. We thus learned that forest cutting regimes, and potentially other factors (e.g., weather and earthworms), have had significant effects on fruit production.
Ironically, black bears are considered a “forest dependent” species, which sometimes leads to the assumption that forest disturbance due to timber harvests must have adverse effects. However, we demonstrated that a landscape with a variety of forest types and age classes tends to be more productive and have a higher diversity of foods.
25
Moreover, different forest types produced a variety of foods at different times of the year, each of which might be available for just a short time (Davis et al., 2006). For example, upland brush sites produced a large flush of biomass; however, most of that was from raspberry, which was only available during a few weeks of the summer. Thus, managing solely for total biomass would not be a prudent strategy. Instead, improved understanding of the complexity of bottom-up pressures is necessary for effective management of bears and other wildlife in these changing landscapes.
26
Tables
Table 1. Species groupings for Generalized Linear Mixed Effects models. Common and scientific name of each bear food sampled during 1980s and 2010s field seasons. Species were also grouped to assess physical and seasonal changes of bear foods.
Summer Short Common Name Scientific Name Fall Foods Tall Shrub Foods Shrub Wild sarsaparilla Aralia nudicaulis X X American Aralia racemosa X X spikenard Currant Ribes spp. X X
Gooseberry Ribes spp. X X
Blueberry Vaccinium spp. X X
Red raspberry Rubus idaeus X X
Blackberry Rubus allegheniensis X X
Red-berried elder Sambucus racemosa X X
Nannyberry Viburnum lentago X X
Juneberry Amelanchier spp. X X
Pin cherry Prunus pensylvanica X X
Choke-cherry Prunus virginiana X X
Wild plum Prunus americana X X Alder-leaved Rhamnus alnifolia X X buckthorn High-bush Viburnum trilobum X X cranberry Downy arrow- Viburnum X X wood rafinesquianum Hawthorn Crataegus spp. X X
American hazel Corylus americana X X
Beaked hazel Corylus cornuta X X
Pagoda dogwood Cornus alternifolia X X Red-osier Cornus sericea X X dogwood Round-leaved Cornus rugosa X X dogwood
27
Table 2. Results of species grouping abundance. Parameter estimates (logit units) from linear mixed model (LMMs; logit-normal) for proportional abundance of summer/fall foods and short/tall shrubs. Reference groups for Decade (2010s), Lowland and Coniferous are 1980s, Upland and Deciduous, respectively.
Abundance (logit-normal) Factor Estimate SE P Summer Food Intercept -4.062 0.178 <0.001* Decade (2010s) -0.384 0.223 0.109 Canopy Cover -0.006 0.002 <0.01* Lowland -0.384 0.130 <0.01* Coniferous 0.052 0.130 0.687
Fall Food Intercept -4.403 0.217 <0.001* Decade (2010s) -0.186 0.248 0.460 Canopy Cover 0.004 0.002 0.131 Lowland -1.384 0.185 <0.001* Coniferous -0.288 0.185 0.121
Short Shrubs Intercept -3.788 0.183 <0.001* Decade (2010s) -0.484 0.229 0.058 Canopy Cover -0.006 0.002 <0.01* Lowland -0.438 0.131 <0.001* Coniferous -0.036 0.130 0.779
Tall Shrubs Intercept -4.554 0.213 <0.001* Decade (2010s) -0.094 0.251 0.713 Canopy Cover 0.003 0.002 0.134 Lowland -1.172 0.178 <0.001* Coniferous -0.297 0.177 0.096 * Significant at P < 0.05
28
Table 3. Productivity results for species groupings. Parameter estimates (logit and log units) from generalized linear mixed model (GLMMs; binomial error, logit link) for productivity ratings and linear mixed models (LMM; log-normal error) for non-zero productivity ratings of summer/fall foods and short/tall shrubs. Reference groups for Decade (2010s), Lowland and Coniferous are 1980s, Upland and Deciduous, respectively.
Fruit Presence (Binomial) Fruit Availability (Log-normal) Factor Estimate SE P Estimate SE P Summer Foods Intercept 3.149 0.460 <0.001* -0.792 0.178 <0.001* Decade (2010s) -2.383 0.350 <0.001* -0.511 0.236 0.056 Canopy Cover -0.021 0.006 <0.001* -0.010 0.002 <0.001* Lowland -0.256 0.307 0.405 0.080 0.111 0.474 Coniferous 0.102 0.300 0.735 0.006 0.102 0.952
Fall Foods Intercept 1.731 0.461 <0.001* -0.751 0.224 <0.01* Decade (2010s) -1.646 0.444 <0.001* -0.505 0.324 0.160 Canopy Cover 0.001 0.006 0.840 -0.003 0.002 0.155 Lowland -2.100 0.390 <0.001* -0.361 0.127 <0.01* Coniferous 0.301 0.334 0.367 0.144 0.103 0.163
Short Shrubs Intercept 3.405 0.461 <0.001* -0.633 0.167 <0.001* Decade (2010s) -2.127 0.325 <0.001* -0.438 0.220 0.078 Canopy Cover -0.023 0.006 <0.001* -0.008 0.002 <0.001* Lowland -0.587 0.309 0.058 -0.168 0.105 0.111 Coniferous -0.096 0.301 0.749 0.042 0.096 0.660
Tall Shrubs Intercept 1.888 0.502 <0.001* -0.912 0.239 <0.01* Decade (2010s) -2.058 0.518 <0.001* -0.571 0.349 0.143 Canopy Cover 0.003 0.007 0.611 -0.005 0.002 <0.05* Lowland -1.873 0.402 <0.001* -0.077 0.127 0.545 Coniferous 0.063 0.350 0.857 0.106 0.106 0.321 * Significant at P < 0.05
29
Table 4. Biomass results for species groupings. Parameter estimates (logit and log units) from generalized linear mixed model (GLMMs; binomial error, logit link) for biomass and linear mixed models (LMM; log- normal error) for non-zero biomass of summer/fall foods and short/tall shrubs. Reference groups for Decade (2010s), Lowland and Coniferous are 1980s, Upland and Deciduous, respectively.
Biomass Availability Biomass Presence (Binomial) (Log-normal) Factor Estimate SE P Estimate SE P Summer Food Intercept 3.109 0.456 <0.001* 1.982 0.378 <0.001*
Decade -2.447 0.345 <0.001* -0.296 0.483 0.553 (2010s) Canopy Cover -0.020 0.006 <0.01* -0.038 0.004 <0.001* Lowland 0.131 0.308 0.670 -0.156 0.266 0.560 Coniferous -0.317 0.313 0.311 -0.145 0.252 0.566
Fall Food Intercept 1.622 0.457 <0.001* 0.552 0.386 0.159 Decade -1.861 0.452 <0.001* -0.396 0.468 0.413 (2010s)
Canopy Cover 0.002 0.006 0.648 -0.006 0.005 0.184 Lowland -2.357 0.403 <0.001* -1.120 0.310 <0.001* Coniferous 0.211 0.337 0.530 -0.576 0.265 <0.05*
Short Shrubs
Intercept 3.234 0.458 <0.001* 0.326 0.370 0.382 Decade -2.213 0.336 <0.001* -0.318 0.457 0.499 (2010s) Canopy Cover -0.021 0.006 <0.001* -0.039 0.005 <0.001* Lowland -0.617 0.317 0.051 -0.867 0.279 <0.01*
Coniferous -0.024 0.310 0.939 -0.205 0.263 0.435
Tall Shrubs
Intercept 1.837 0.513 <0.001* -2.269 0.408 <0.001* Decade -2.210 0.542 <0.001* -0.616 0.538 0.282 (2010s) Canopy Cover 0.005 0.007 0.467 0.001 0.005 0.913 Lowland -2.167 0.428 <0.001* -0.269 0.266 0.315
Coniferous -0.094 0.364 0.796 -0.492 0.229 <0.05*
* Significant at P < 0.05
30
Figures
Figure 1. Map of study area (red box). Different public and private land ownership (colors) were distributed across the landscape. The study area contains numerous small to medium size lakes with public access.
31
Figure 2. Mean species abundance (%) for each forest stand type for the 1980s and 2010s sampling periods. Species with a mean abundance < 3% were grouped as “other” to improve clarity. Forest type abbreviations: ASH=black ash, ASP=mature aspen (>30 yrs), BIR=birch, CED=white cedar, LAS=lowland aspen, LBR=lowland brush, MAP=maple, PIN=red pine (>35 yrs), PLA=pine plantation (8-20 yrs), REG=aspen regeneration (5-15 yrs), SPR=spruce-tamarack, UBR=upland brush/clearcuts.
32
Figure 3. Probability of non-zero productivity by sampling period. Probability of a forest stand producing any amount of food (with 95% bootstrapped prediction intervals) declined for all species groups from the 1980s to the 2010s.
33
Figure 4. Probability of non-zero productivity in relation to canopy cover. Probability of short shrubs and summer foods having a productivity rating greater than zero (with 95% bootstrapped prediction intervals) declined as canopy cover increased (P < 0.001). The difference in probability between the 1980s and 2010s was significant (P < 0.001).
34
Figure 5. Year-to-year fluctuations in summer food productivity. Yearly mean productivity ratings (on a 0- 4 scale) for common summer foods that occurred in common and productive stand types. Year-to-year fluctuations for specific species do not always follow the same pattern in different forest types. The magnitude of fluctuations were greater during the 1980s than the 2010s. Forest abbreviations are the same as Figure 2.
35
Figure 6. Year-to-year fluctuations in fall food productivity. Yearly mean productivity ratings (on a 0-4 scale) for common fall foods that occurred in common and productive stand types. Year-to-year fluctuations for specific species do not always follow the same pattern in different forest types. The magnitude of fluctuations were greater during the 1980s than the 2010s, yet were more comparable than summer foods. Forest abbreviations are the same as Figure 2.
36
Figure 7. Mean species biomass by sampling period. Comparison of sampling period biomass (kg/ha) by species in each forest stand type. Species that produced on average < 5 kg/ha were grouped into "other" to help with clarity. Forest abbreviations are the same as Figure 2.
37
Figure 8. Non-zero biomass in relation to canopy cover. When biomass was present in a forest stand, short shrub and summer food biomass (with 95% bootstrapped prediction intervals) declined as canopy cover increased. Summer food biomass was greater, albeit more variable, at lower canopy cover estimates than short shrubs.
38
Figure 9. Canopy cover estimates by sampling period. Comparison of sampling period mean canopy cover estimates (%) for each forest stand type. Error bars depict 95% bootstrapped confidence intervals. Canopy cover significantly increased in mature aspen (>30 yrs) black ash, lowland aspen (mixed with balsam poplar), and mature red pine (>35 yrs) and declined in upland brush (P < 0.01). Forest abbreviations are the same as Figure 2.
39
Figure 10. Landscape biomass by sampling period. Landscape biomass (kg), and standard error, in relation to the forest type availability on the landscape. Forest type availability was estimated using habitat information collected during telemetry flights in the late 1980s and stand inventory data from US Forest Service, Minnesota Department of Natural Resources and the Itasca county Land Department for the 2010s. Forest abbreviations are the same as Figure 2.
40
Figure 11. Landscape biomass by landownership. Landscape biomass (kg), and standard error, in relation to the forest type composition on different public land ownerships. For this comparison, proportions of each habitat type were calculated using stand inventory data from US Forest Service, Minnesota Department of Natural Resources and the Itasca county Land Department for the 2010s. The area between federal vs state/county was assumed to be the same (total area of public land = 36,692 ha) and the proportions of habitat types were multiplied by this area. Forest abbreviations are the same as Figure 2.
41
References
Ågren, J., 1988. Between-year variation in flowering and fruit set in frost-prone and frost- sheltered populations of dioecious Rubus chamaemorus. Oecologia 76, 175–183. https://doi.org/10.1007/BF00379950 Ahlering, M.A., Faaborg, J., 2006. Avian habitat management meets conspecific attraction: if you build it, will they come? The Auk 123, 301–312. https://doi.org/10.1642/0004- 8038(2006)123[301:AHMMCA]2.0.CO;2 Akaike, H., 1987. Factor Analysis and AIC, in: Selected Papers of Hirotugu Akaike, Springer Series in Statistics. Springer, New York, NY, pp. 371–386. https://doi.org/10.1007/978- 1-4612-1694-0_29 Allen, C.D., Macalady, A.K., Chenchouni, H., Bachelet, D., McDowell, N., Vennetier, M., Kitzberger, T., Rigling, A., Breshears, D.D., Hogg, E.H. (Ted), Gonzalez, P., Fensham, R., Zhang, Z., Castro, J., Demidova, N., Lim, J.-H., Allard, G., Running, S.W., Semerci, A., Cobb, N., 2010. A global overview of drought and heat-induced tree mortality reveals emerging climate change risks for forests. For. Ecol. Manag., Adaptation of Forests and Forest Management to Changing Climate 259, 660–684. https://doi.org/10.1016/j.foreco.2009.09.001 Allendorf, F.W., England, P.R., Luikart, G., Ritchie, P.A., Ryman, N., 2008. Genetic effects of harvest on wild animal populations. Trends Ecol. Evol. 23, 327–337. https://doi.org/10.1016/j.tree.2008.02.008 Andrén, H., 1994. Effects of Habitat Fragmentation on Birds and Mammals in Landscapes with Different Proportions of Suitable Habitat: A Review. Oikos 71, 355–366. https://doi.org/10.2307/3545823 Arnold, T.W., 2010. Uninformative Parameters and Model Selection Using Akaike’s Information Criterion. J. Wildl. Manag. 74, 1175–1178. https://doi.org/10.2193/2009-367 Ayres, M.P., Lombardero, M.J., 2000. Assessing the consequences of global change for forest disturbance from herbivores and pathogens. Sci. Total Environ., Climate change,Forests and 262, 263–286. https://doi.org/10.1016/S0048-9697(00)00528-3 Bates, D., Mächler, M., Bolker, B., Walker, S., 2014. Fitting Linear Mixed-Effects Models using lme4. ArXiv14065823 Stat. Beckmann, J.P., Berger, J., 2003. Rapid ecological and behavioural changes in carnivores: the responses of black bears (Ursus americanus) to altered food. J. Zool. 261, 207–212. https://doi.org/10.1017/S0952836903004126 Bohlen, P.J., Scheu, S., Hale, C.M., McLean, M.A., Migge, S., Groffman, P.M., Parkinson, D., 2004. Non‐native invasive earthworms as agents of change in northern temperate forests. Front. Ecol. Environ. 2, 427–435. https://doi.org/10.1890/1540- 9295(2004)002[0427:NIEAAO]2.0.CO;2 Bowlby, J.N., Roff, J.C., 1986. Trophic Structure in Southern Ontario Streams. Ecology 67, 1670–1679. https://doi.org/10.2307/1939099 Bridges, A.S., Vaughan, M.R., Fox, J.A., 2011. Reproductive Ecology of American Black Bears in the Alleghany Mountains of Virginia, USA. J. Wildl. Manag. 75, 1137–1144. https://doi.org/10.1002/jwmg.148 Bro, E., Mayot, P., Corda, E., Reitz, F., 2004. Impact of habitat management on grey partridge populations: assessing wildlife cover using a multisite BACI experiment. J. Appl. Ecol. 41, 846–857. https://doi.org/10.1111/j.0021-8901.2004.00939.x Burnham, K.P., Anderson, D.R., 2004. Multimodel Inference: Understanding AIC and BIC in Model Selection. Sociol. Methods Res. 33, 261–304. https://doi.org/10.1177/0049124104268644
42
Clark, J.D., Clapp, D.L., Smith, K.G., Ederington, B., 1994. Black Bear Habitat Use in Relation to Food Availability in the Interior Highlands of Arkansas. Bears Their Biol. Manag. 9, 309–318. https://doi.org/10.2307/3872716 Clark, J.D., Van Manen, F.T., Pelton, M.R., Morrison, 2005. Bait stations, hard mast, and black bear population growth in great smoky mountains national park. J. Wildl. Manag. 69, 1633–1640. https://doi.org/10.2193/0022-541X(2005)69[1633:BSHMAB]2.0.CO;2 Cleland, E.E., Chuine, I., Menzel, A., Mooney, H.A., Schwartz, M.D., 2007. Shifting plant phenology in response to global change. Trends Ecol. Evol. 22, 357–365. https://doi.org/10.1016/j.tree.2007.04.003 Coffin, A.W., 2007. From roadkill to road ecology: A review of the ecological effects of roads. J. Transp. Geogr. 15, 396–406. https://doi.org/10.1016/j.jtrangeo.2006.11.006 Collins, C., Kays, R., 2011. Causes of mortality in North American populations of large and medium-sized mammals. Anim. Conserv. 14, 474–483. https://doi.org/10.1111/j.1469- 1795.2011.00458.x Costello, C.M., Jones, D.E., Inman, R.M., Inman, K.H., Thompson, B.C., Quigley, H.B., 2003. Relationship of Variable Mast Production to American Black Bear Reproductive Parameters in New Mexico. Ursus 14, 1–16. Costello, C.M., Sage, R.W., Jr., 1994. Predicting Black Bear Habitat Selection from Food Abundance under 3 Forest Management Systems. Bears Their Biol. Manag. 9, 375–387. https://doi.org/10.2307/3872724 Dale, V.H., Joyce, L.A., Mcnulty, S., Neilson, R.P., Ayres, M.P., Flannigan, M.D., Hanson, P.J., Irland, L.C., Lugo, A.E., Peterson, C.J., Simberloff, D., Swanson, F.J., Stocks, B.J., Michael Wotton, B., 2001. Climate Change and Forest Disturbances. BioScience 51, 723–734. https://doi.org/10.1641/0006-3568(2001)051[0723:CCAFD]2.0.CO;2 Davis, H., Weir, R.D., Hamilton, A.N., Deal, J.A., 2006. Influence of phenology on site selection by female American black bears in coastal British Columbia. Ursus 17, 41–51. https://doi.org/10.2192/1537-6176(2006)17[41:IOPOSS]2.0.CO;2 Ditmer, M.A., Garshelis, D.L., Noyce, K.V., Haveles, A.W., Fieberg, J.R., 2016. Are American black bears in an agricultural landscape being sustained by crops? J. Mammal. 97, 54–67. https://doi.org/10.1093/jmammal/gyv153 Dobson, A.M., Blossey, B., Richardson, J.B., 2017. Invasive earthworms change nutrient availability and uptake by forest understory plants. Plant Soil 421, 175–190. https://doi.org/10.1007/s11104-017-3412-9 Dukes, J.S., Pontius, J., Orwig, D., Garnas, J.R., Rodgers, V.L., Brazee, N., Cooke, B., Theoharides, K.A., Stange, E.E., Harrington, R., Ehrenfeld, J., Gurevitch, J., Lerdau, M., Stinson, K., Wick, R., Ayres, M., 2009. Responses of insect pests, pathogens, and invasive plant species to climate change in the forests of northeastern North America: What can we predict?This article is one of a selection of papers from NE Forests 2100: A Synthesis of Climate Change Impacts on Forests of the Northeastern US and Eastern Canada. Can. J. For. Res. 39, 231–248. https://doi.org/10.1139/X08-171 Efron, B., Tibshirani, R.J., 1994. Hypothesis testing with the bootstrap, in: An Introduction to the Bootstrap, Monographs on Statistics and Applied Probability. CRC Press, pp. 220–236. Eisenhauer, N., Stefanski, A., Fisichelli, N.A., Rice, K., Rich, R., Reich, P.B., 2014. Warming shifts ‘worming’: effects of experimental warming on invasive earthworms in northern North America. Sci. Rep. 4, 6890. https://doi.org/10.1038/srep06890 Elowe, K.D., Dodge, W.E., 1989. Factors Affecting Black Bear Reproductive Success and Cub Survival. J. Wildl. Manag. 53, 962–968. https://doi.org/10.2307/3809596 Estes, J.A., 1995. Top-Level Carnivores and Ecosystem Effects: Questions and Approaches, in: Linking Species & Ecosystems. Springer, Boston, MA, pp. 151–158. https://doi.org/10.1007/978-1-4615-1773-3_15
43
Frederiksen, M., Edwards, M., Richardson, A.J., Halliday, N.C., Wanless, S., 2006. From plankton to top predators: bottom-up control of a marine food web across four trophic levels. J. Anim. Ecol. 75, 1259–1268. https://doi.org/10.1111/j.1365-2656.2006.01148.x Frelich, L.E., Hale, C.M., Reich, P.B., Holdsworth, A.R., Scheu, S., Heneghan, L., Bohlen, P.J., 2006. Earthworm invasion into previously earthworm-free temperate and boreal forests, in: Biological Invasions Belowground: Earthworms as Invasive Species. Springer, Dordrecht, pp. 35–45. https://doi.org/10.1007/978-1-4020-5429-7_5 Gamelon, M., Besnard, A., Gaillard, J.-M., Servanty, S., Baubet, E., Brandt, S., Gimenez, O., 2011. High Hunting Pressure Selects for Earlier Birth Date: Wild Boar as a Case Study. Evolution 65, 3100–3112. https://doi.org/10.1111/j.1558-5646.2011.01366.x Garner, J.D., Nelson, M.D., Tavernia, B.G., Housman, I.W., Perry, C.H., 2016. Early successional forest and land cover geospatial data of the upper Midwest. U.S. Forest Service, Fort Collins, CO. Garshelis, D.L., Noyce, K.V., 2008. Seeing the world through the nose of a bear-diversity of foods fosters behavioral and demographic stability, in: Fullbright, T.E., Hewitt, D.G. (Eds.), Wildlife Science. Linking Ecological Theory and Management Applications. CRC Presss, Boca Raton, Florida, USA, pp. 139–163. Garshelis, D.L., Noyce, K.V., Coy, P.L., 1998. Calculating Average Age of First Reproduction Free of the Biases Prevalent in Bear Studies. Ursus 10, 437–447. Garshelis, D.L., Scheick, B.K., Doan-Crider, D.L., Beecham, J.J., Obbard, M.E., 2016. Ursus americanus. https://doi.org/10.2305/IUCN.UK.2016-3.RLTS.T41687A45034604.en Garshelis, D.L., Tri, A., 2018. Status of Minnesota black bears, 2017 (Final Report to Bear Committee). Minnesota Department of Natural Resources. Gripenberg, S., Roslin, T., 2007. Up or down in space? Uniting the bottom-up versus top-down paradigm and spatial ecology. Oikos 116, 181–188. https://doi.org/10.1111/j.0030- 1299.2007.15266.x Hale, C.M., Frelich, L.E., Reich, P.B., 2006. Changes in Hardwood Forest Understory Plant Communities in Response to European Earthworm Invasions. Ecology 87, 1637–1649. https://doi.org/10.1890/0012-9658(2006)87[1637:CIHFUP]2.0.CO;2 Hall, D.B., 2000. Zero‐Inflated Poisson and Binomial Regression with Random Effects: A Case Study. Biometrics 56, 1030–1039. https://doi.org/10.1111/j.0006-341X.2000.01030.x Hall, Motzkin, G., Foster, D.R., Syfert, M., Burk, J., 2003. Three hundred years of forest and land‐use change in Massachusetts, USA. J. Biogeogr. 29, 1319–1335. https://doi.org/10.1046/j.1365-2699.2002.00790.x Hashimoto, Y., Kaji, M., Sawada, H., Takatsuki, S., 2003. Five-year study on the autumn food habits of the Asiatic black bear in relation to nut production. Ecol. Res. 18, 485–492. https://doi.org/10.1046/j.1440-1703.2003.00572.x Heinselman, M.L., 1954. The Extent of Natural Conversion to Other Species in the Lake-States Aspen-Birch Type. J. For. 52, 737–738. https://doi.org/10.1093/jof/52.10.737 Hertel, A.G., Bischof, R., Langval, O., Mysterud, A., Kindberg, J., Swenson, J.E., Zedrosser, A., 2017. Berry production drives bottom–up effects on body mass and reproductive success in an omnivore. Oikos n/a-n/a. https://doi.org/10.1111/oik.04515 Hertel, A.G., Steyaert, S.M.J.G., Zedrosser, A., Mysterud, A., Lodberg-Holm, H.K., Gelink, H.W., Kindberg, J., Swenson, J.E., 2016. Bears and berries: species-specific selective foraging on a patchily distributed food resource in a human-altered landscape. Behav. Ecol. Sociobiol. 70, 831–842. https://doi.org/10.1007/s00265-016-2106-2 Hogg, E.H. (Ted), Bernier, P.Y., 2005. Climate change impacts on drought-prone forests in western Canada. For. Chron. 81, 675–682. https://doi.org/10.5558/tfc81675-5 Holdsworth, A.R., Frelich, L.E., Reich, P.B., 2007a. Regional extent of an ecosystem engineer: earthworm invasion in northern hardwood forests. Ecol. Appl. 17, 1666–1677. https://doi.org/10.1890/05-2003.1
44
Holdsworth, A.R., Frelich, L.E., Reich, P.B., 2007b. Effects of Earthworm Invasion on Plant Species Richness in Northern Hardwood Forests. Conserv. Biol. 21, 997–1008. https://doi.org/10.1111/j.1523-1739.2007.00740.x Howe, E.J., Obbard, M.E., Bowman, J., 2012. Prior reproduction and weather affect berry crops in central Ontario, Canada. Popul. Ecol. 54, 347–356. https://doi.org/10.1007/s10144- 011-0301-6 Huang, J.-G., Stadt, K.J., Dawson, A., Comeau, P.G., 2013. Modelling Growth-Competition Relationships in Trembling Aspen and White Spruce Mixed Boreal Forests of Western Canada. PLoS ONE 8. https://doi.org/10.1371/journal.pone.0077607 Hunter, M.D., Price, P.W., 1992. Playing Chutes and Ladders: Heterogeneity and the Relative Roles of Bottom-Up and Top-Down Forces in Natural Communities. Ecology 73, 724– 732. https://doi.org/10.2307/1940152 Inouye, D.W., 2008. Effects of climate change on phenology, frost damage, and floral abundance of montane wildflowers. Ecology 89, 353–362. https://doi.org/10.1890/06-2128.1 Inouye, D.W., 2001. The ecological and evolutionary significance of frost in the context of climate change. Ecol. Lett. 3, 457–463. https://doi.org/10.1046/j.1461- 0248.2000.00165.x Itasca County, 2018. Forest Management | Itasca County, MN [WWW Document]. Itasca Cty. Minn. URL https://www.co.itasca.mn.us/234/Forest-Management (accessed 4.10.18). Kavčič, I., Adamič, M., Kaczensky, P., Krofel, M., Kobal, M., Jerina, K., 2015. Fast food bears: brown bear diet in a human-dominated landscape with intensive supplemental feeding. Wildl. Biol. 21, 1–8. https://doi.org/10.2981/wlb.00013 Keyser, A.R., Kimball, J.S., Nemani, R.R., Running, S.W., 2000. Simulating the effects of climate change on the carbon balance of North American high-latitude forests. Glob. Change Biol. 6, 185–195. https://doi.org/10.1046/j.1365-2486.2000.06020.x Kontio, B.D., Garshelis, D.L., Birney, E.C., Andersen, D.E., 1998. Resilience of a Minnesota Black Bear Population to Heavy Hunting: Self-Sustaining Population or Population Sink? Ursus 10, 139–146. Kramer, K., Leinonen, I., Loustau, D., 2000. The importance of phenology for the evaluation of impact of climate change on growth of boreal, temperate and Mediterranean forests ecosystems: an overview. Int. J. Biometeorol. 44, 67–75. https://doi.org/10.1007/s004840000066 Kurz, W.A., Dymond, C.C., Stinson, G., Rampley, G.J., Neilson, E.T., Carroll, A.L., Ebata, T., Safranyik, L., 2008. Mountain pine beetle and forest carbon feedback to climate change. Nature 452, 987–990. https://doi.org/10.1038/nature06777 Lamb, C.T., Mowat, G., McLellan, B.N., Nielsen, S.E., Boutin, S., 2017. Forbidden fruit: human settlement and abundant fruit create an ecological trap for an apex omnivore. J. Anim. Ecol. 86, 55–65. https://doi.org/10.1111/1365-2656.12589 Laundré, J.W., Hernández, L., Medina, P.L., Campanella, A., López-Portillo, J., González- Romero, A., Grajales-Tam, K.M., Burke, A.M., Gronemeyer, P., Browning, D.M., 2014. The landscape of fear: the missing link to understand top-down and bottom-up controls of prey abundance? Ecology 95, 1141–1152. Lechowicz, M.J., 1995. Seasonality of flowering and fruiting in temperate forest trees. Can. J. Bot. 73, 175–182. https://doi.org/10.1139/b95-021 Leopold, A., 1933. The Conservation Ethic. J. For. 31, 634–643. https://doi.org/10.1093/jof/31.6.634 McLaughlin, C.R., Matula, G.J., O’Connor, R.J., 1994. Synchronous Reproduction by Maine Black Bears. Bears Their Biol. Manag. 9, 471–479. https://doi.org/10.2307/3872735 Meserve, P.L., Kelt, D.A., Milstead, W.B., Gutiérrez, J.R., 2003. Thirteen Years of Shifting Top- Down and Bottom-Up Control. BioScience 53, 633–646. https://doi.org/10.1641/0006- 3568(2003)053[0633:TYOSTA]2.0.CO;2
45
Miller, T.E.X., 2008. Bottom-up, top-down, and within-trophic level pressures on a cactus- feeding insect. Ecol. Entomol. 33, 261–268. https://doi.org/10.1111/j.1365- 2311.2007.00964.x Milner, J.M., Nilsen, E.B., Andreassen, H.P., 2007. Demographic Side Effects of Selective Hunting in Ungulates and Carnivores. Conserv. Biol. 21, 36–47. https://doi.org/10.1111/j.1523-1739.2006.00591.x Minnesota Department of Natural Resources, 2018. MNDNR Forest Stand Inventory - Minnesota Geospatial Commons [WWW Document]. Minn. Geospatial Commons. URL https://gisdata.mn.gov/dataset/biota-dnr-forest-stand-inventory (accessed 4.10.18). NOAA, 2018. Climate Data Online (CDO) - The National Climatic Data Center’s (NCDC) [WWW Document]. NOAA Natl. Cent. Environemtnal Inf. URL https://www.ncdc.noaa.gov/cdo-web/ (accessed 4.11.18). Noyce, K.V., Coy, P.L., 1990. Abundance and Productivity of Bear Food Species in Different Forest Types of Northcentral Minnesota. Bears Their Biol. Manag. 8, 169–181. https://doi.org/10.2307/3872917 Noyce, K.V., Garshelis, D.L., 2014. Follow the leader: social cues help guide landscape-level movements of American black bears (Ursus americanus). Can. J. Zool. 92, 1005–1017. https://doi.org/10.1139/cjz-2014-0029 Noyce, K.V., Garshelis, D.L., 2011. Seasonal migrations of black bears (
46
United States Forest Service, 2018. Chippewa National Forest - Geospatial Data [WWW Document]. URL https://www.fs.usda.gov/main/chippewa/landmanagement/gis (accessed 4.10.18). United States Forest Service, 2017. Chippewa National Forest Timber Statistics (Annual Timber Program Statistics). USDA. United States Forest Service, 2018. Chippewa National Forest - About the Forest [WWW Document]. Chippewa Natl. For. For. URL https://www.fs.usda.gov/main/chippewa/about-forest (accessed 4.22.18). Verbeke, G., 1997. Linear Mixed Models for Longitudinal Data, in: Linear Mixed Models in Practice, Lecture Notes in Statistics. Springer, New York, NY, pp. 63–153. https://doi.org/10.1007/978-1-4612-2294-1_3 Volney, W.J.A., Fleming, R.A., 2000. Climate change and impacts of boreal forest insects. Agric. Ecosyst. Environ. 82, 283–294. https://doi.org/10.1016/S0167-8809(00)00232-2 Vospernik, S., Reimoser, S., 2008. Modelling changes in roe deer habitat in response to forest management. For. Ecol. Manag. 255, 530–545. https://doi.org/10.1016/j.foreco.2007.09.036 Warton, D.I., Hui, F.K.C., 2011. The arcsine is asinine: the analysis of proportions in ecology. Ecology 92, 3–10. https://doi.org/10.1890/10-0340.1
47
Appendix
Appendix Figure 1. Mean abundance estimates for fall foods, short shrubs, and tall shrubs in lowland and upland forest types with 95% bootstrapped prediction intervals. These species group abundances were significantly lower in lowland forest type than in upland forest types (P < 0.001).
48
Appendix Figure 2. Short shrub abundance (%) in relation to canopy cover with 95% bootstrapped prediction intervals. Short shrub abundance declined as canopy cover increased (P < 0.001).
49
Appendix Figure 3. Probability of fall food and tall shrubs have a productivity rating greater than zero (with 95% bootstrapped prediction intervals) was lower in lowland forests than upland forests.
50
Appendix Figure 4. When fruits were present, the productivity rating (with 95% bootstrapped prediction intervals) for short shrubs and summer foods declined as canopy cover increased (P < 0.001).
51
Appendix Figure 5. Probability of a forest stand producing any biomass greater than zero (with 95% bootstrapped prediction intervals) decreased for all species groups from 1980s to the 2010s (P < 0.001).
52
Appendix Figure 6. The probability of a stand having short shrub or summer food biomass greater than zero (with 95% bootstrapped prediction intervals) declined as canopy cover increased (P < 0.001)
53
Appendix Figure 7. The estimated proportion of each forest type from the 1980s to the 2010s across what? Most mature forests, especially mature aspen, substantially increased in availability between the 1980s and 2010s. The area of young forests decreased over time with the exception of upland brush. Forest abbreviations are the same as Figure 2.
54
Appendix Figure 8. Comparison of the estimated versus true proportion of habitat types used to estimate landscape biomass in the 2010s. Mature aspen, spruce/tamarack and upland brush were slightly overrepresented, whereas, ash, birch and maple, mature pine, and aspen regeneration were slightly 2 underrepresented (X 11=1836.3, P < 0.001). Lowland brush was not used in this comparison because there is no reliable spatial layer for this habitat type on all public lands. Forest abbreviations are the same as Figure 2.
55
Appendix Figure 9. Comparison of federal versus state/county of habitat type proportions in the 2010s. Lowland brush was not used in this comparison because there is no reliable spatial layer for this habitat type on all public lands. Forest abbreviations are the same as Figure 2.
56
Appendix Table 1. Predicted flowering dates (using the median accumulated growing degree days) and the number of potential frost events after the flowering date. Potential frost events were defined as days were the minimum temperature dropped below freezing (0° Celsius).
Number of Year Species Flowering Date Potential Frost Events
1984 Chokecherry May 18th 2
Red-osier Dogwood June 10th 0
1985 Chokecherry May 10th 2
Red-osier Dogwood June 7th 0
1986 Chokecherry May 13th 3
Red-osier Dogwood June 7th 0
1987 Chokecherry May 2nd 2
Red-osier Dogwood May 30th 0
1988 Chokecherry May 15th 1
Red-osier Dogwood June 5th 0
1989 Chokecherry May 21st 1
Red-osier Dogwood June 17th 0
2015 Chokecherry May 15th 3
Red-osier Dogwood June 10th 0
2016 Chokecherry May 9th 4
Red-osier Dogwood June 6th 0
2017 Chokecherry May 11th 0
Red-osier Dogwood June 8th 0
57